impact of interference between neighbouring 5g micro...
TRANSCRIPT
Impact of Interference Between Neighbouring 5G Micro
Operators
Kimmo Hiltunen, Marja Matinmikko-Blue, Matti Latva-aho
Centre for Wireless Communications (CWC), University of Oulu, Oulu, Finland
Corresponding author: Kimmo Hiltunen, [email protected], +358504018897
Abstract: Local small cell deployments will become an important part of the future 5G networks, in particular
in the higher frequency bands. In order to speed up the wide-spread deployment of such ultra-dense
networks, new business and spectrum authorization models are needed. The recently proposed concept of
micro operators with local spectrum micro licensing has gained significant interest in research, industry and
regulation to complement the traditional models based on networks deployed and operated by the mobile
network operators (MNOs). While assessing the applicability of the proposed micro operator concept, one
important aspect is to evaluate the impact of the inter-operator interference on the performance of the
victim network when deployed in the same or adjacent channel. To support such interference evaluations
between micro operators, this paper proposes a deployment scenario including two neighbouring buildings,
propagation models for connections both within a building and between the buildings, and a criteria for the
required minimum separation distance based on the observed throughput loss. Finally, system simulations
are performed to evaluate the impact of the key deployment aspects on the required minimum separation
distance between the micro operators in the 3.5 GHz band. The obtained results indicate that the required
minimum separation distances are highly scenario-specific, which needs to be considered in the overall local
spectrum micro licensing model development and the setup of appropriate rules to coordinate the
interference.
Keywords: 5G; micro operator; spectrum sharing; radio wave propagation; interference management; radio
network performance;
Acknowledgments: Authors would like to acknowledge Business Finland for funding the “Micro-operator
concept for boosting local service delivery in 5G (uO5G)” project.
1 Introduction Next generation mobile communication networks known as 5G are expected to open prospects for wide-
spread digitalization and new innovative business models by forming the ground for high quality mobile
broadband and new types of high-demand applications. Provisioning of high quality wireless connectivity in
specific high-demand areas such as campuses, transport hubs, public service providers’ units and enterprises
has become a key societal objective in Europe as the enabler for new services [1]. Customized access for
different stakeholders has become a major design criteria in the development of 5G networks and their
applications in different vertical sectors to address location specific needs for wireless connectivity [2].
Furthermore, the growing need for locally operated wireless network deployments by different stakeholders
has become a reality in 5G to meet the ever increasing requirements for higher capacity, higher data rate,
lower latency, massive device density, and reduced capital and operational costs [3-5].
Deployment of new mobile communication networks is highly dependent on the timely availability of 5G
spectrum which in turn is dependent on the regulatory decisions. While the 5G networks are expected to be
deployed in a wide range of frequency bands ranging from below 1 GHz and between 1-6 GHz, major efforts
are spent on making new spectrum bands available above 6 GHz especially in the millimetre wave range (24-
86 GHz). With the global commitment to make new 5G spectrum available the debate on suitable spectrum
authorization models to gain access to 5G spectrum has started in the regulatory domain by addressing a mix
of licensed and unlicensed models [1, 6]. The regulatory community has recently adopted several sharing-
based spectrum authorization models to allow additional users in the band while protecting the incumbent
spectrum users. Recent European developments include a two-tier Licensed Shared Access (LSA) model
which is discussed for allowing mobile communication networks to use the 2.3-2.4 GHz and 3.6-3.8 GHz bands
while protecting the existing incumbents [7]. In the US a three-tier sharing model provides Citizens
Broadband Radio Service (CBRS) by introducing two layers of additional users in local areas under licensed
and general authorization regimes in the 3.55-3.7 GHz band while protecting the incumbents [8]. These
sharing-based models become increasingly important for the deployment of 5G networks.
The research community has started to address new 5G deployments in the millimetre wave bands where
the challenge is to achieve sufficient indoor coverage and capacity. At lower frequency bands, general indoor
coverage can be obtained with outdoor base stations, while the local high capacity needs can be satisfied by
deploying dedicated in-building solutions, such as distributed antenna systems and pico base stations, within
the traffic hotspots. The problem with these kind of operator-deployed in-building solutions is that they do
not scale so well when the number of buildings requiring in-building solutions is increasing, either due to
increased capacity requirements or due to increased center frequencies. Instead of relying solely on the
networks deployed separately by each of the mobile network operators (MNOs), a more efficient solution
could be to allow also special micro operators, for example the venue owners, to deploy and operate their
own in-building solutions in order to satisfy local capacity and coverage needs [4]. This kind of micro operator
concept [3] with local spectrum micro licensing model [9, 10] has been recently introduced for local 5G
network deployments for serving its own restricted customer set and acting as neutral host for the customers
of the overlying MNOs [11]. Further discussion on sharing-based based spectrum access models for
millimetre wave bands can be found in [12] and [13] for uncoordinated and restricted sharing, respectively.
The recently proposed locally issued spectrum micro licenses [9, 10] require some form of interference
coordination between the different license holders and potential incumbent spectrum users in the band to
guarantee that their operations are within the predefined interference conditions. For the development of
such interference coordination mechanisms, interference characterization between the involved systems is
critical. Therefore, proper modelling of the radio wave propagation characteristics in the specific frequency
bands and deployment scenarios is required. An initial analysis of the interference between adjacent micro
operator deployments in 3.5 GHz was presented in [9] where the minimum separation distances were
calculated between an interfering base station and a victim mobile terminal based on a fixed interference
criterion. This paper continues the work in [9] by enhancing the models for the micro operator deployment
scenario, radio wave propagation and the criterion for harmful interference to consider more realistic
deployments.
The main contributions of this paper include the following:
We introduce a novel deployment scenario for modelling the interference between micro operators located within neighbouring buildings. To the best of our knowledge, modelling of the building-to-building wave propagation has not been widely discussed in the literature. One such model, the so-called 3GPP dual-strip model, has been introduced in [14]. However, the 3GPP dual-strip model has its weaknesses: it is tailored for 2 GHz or 3.5 GHz and is not valid for higher frequencies, the building penetration model considers only one of the building walls, the impact of both the wall material and the angle of incidence on the penetration loss is not taken into account, and the outdoor path loss between the buildings is to a large extent based on a macro cellular non-line-of-sight model even at relatively short distances. The model proposed in this paper aims to consider also these aspects when modelling the path losses between the neighbouring micro operators.
We develop a new criteria for determining the minimum separation distance between the micro operators. Traditionally, the minimum separation distance is based on an absolute interference threshold, for example 6 dB below the thermal noise floor of the victim receiver. In this paper, we propose a new kind of approach, where the impact on the performance of the victim system is evaluated instead. The aim of the new method is to obtain a more realistic view on the actual interference situation, and in that way to make the reuse of spectrum more efficient.
Finally, we investigate the impact of building wall material, spectrum reuse, wave propagation conditions (line-of-sight versus non-line-of-sight) and victim network deployment on the required minimum distance between the neighbouring micro operators.
The rest of this paper is organized as follows. In Section 2 the concept of local micro operators is introduced
and the different inter-operator interference scenarios are discussed. Section 3 describes the proposed
system model including the assumed network layout, building-to-building propagation model and a model
for the user performance. Simulation results evaluating the impact of inter-operator interference are
presented and analysed in Section 4. Finally, the obtained results are summarized in Section 5 and a few
additional topics for future research are discussed.
2 Local 5G micro operator deployments It is expected that 5G will introduce very dense small cell deployments in specific locations to serve the
versatile needs of different vertical sectors [2]. There is a growing interest for different stakeholders to deploy
these local cellular networks in addition to the currently dominant MNOs as discussed in [4]. In this section
the new micro operator concept [3] is presented and the resulting inter-operator interference scenarios are
characterized.
2.1 Micro operator concept The spectrum authorization models to assign spectrum access rights to those who are requesting them will
shape the future 5G mobile communication market. Traditional spectrum authorization models for providing
mobile services including individual access rights typically obtained through auctions have led to a small
number of MNOs to deploy nation-wide networks with high infrastructure investments. Currently, the only
option for non-MNOs to deploy local networks is through the general authorization (unlicensed) model for
the establishment of wireless local area networks (WLAN) without quality guarantees. These spectrum
authorization models will need to be rethought in 5G as the networks are envisaged to operate also in
considerably higher frequency bands with smaller coverage areas while fulfilling more stringent performance
requirements, which calls for the development of new sharing-based spectrum authorization models. An
unlicensed model where the facility owner would deploy local ultra-dense networks was discussed in [4].
Alternative network deployment models by different stakeholders were presented in [5] to make internet
access affordable for all.
The concept of micro operators was recently proposed in [3] to allow different stakeholders to exploit their
domain specific knowledge and establish locally operated small cell networks in various places such as
shopping malls, hospitals, sports arenas, campuses and enterprises based on local spectrum availability.
Motivated by the need for high-quality guaranteed wireless connectivity in these different deployment areas,
the micro operators were proposed to obtain spectrum micro licenses with a predefined level of local
exclusivity in [9, 10] to deploy and operate small cell networks in a specific location for a given license
duration. The local spectrum micro licensing model combines the benefits of both exclusive licensing and
unlicensed models by allowing a larger number of stakeholders to get quality-guaranteed local spectrum
access rights. The micro licensing model has the potential to become a key enabler to new entry to the mobile
communication market by allowing various stakeholders to become micro operators. With local spectrum
availability they can deploy and operate 5G small cell networks in specific high-demand areas for tailored
service delivery to complement MNO offerings to realize the 5G deployment plans set for example in [6].
2.2 Interference between neighbouring micro operators Spectrum for 5G networks is planned to be made available from various frequency ranges including existing
bands below 1 GHz and between 1-6 GHz, and particularly the new bands above 6 GHz [6]. For the regulators
to allow 5G mobile communication networks to be deployed in these bands, there is a need for careful
sharing and co-existence studies to ensure the feasibility of spectrum sharing between entrant 5G networks
and incumbent systems. In addition, sharing and coexistence between different entrant system deployments
in co-channel and adjacent channel cases need to be studied. The recently proposed locally issued spectrum
micro licenses in [9, 10] will need to define the level of protection from interference which results in the need
for some form of interference coordination between the different license holders and potential incumbent
spectrum users.
In order to study the interference between local micro license holders, a reasonable assumption is that while
the base stations within a micro operator network are coordinated (or synchronized), the neighbouring base
stations belonging to different micro operators will be uncoordinated (or unsynchronized). In case of dynamic
time division duplex (TDD) [15], a synchronized deployment means that the neighbouring cells have the same
uplink:downlink (UL:DL) ratio and that the uplink slots in one cell are always aligned with the uplink slots in
the other cells. Hence, in case of small cell deployments, there will never be time instants when the uplink
slots in one cell will be interfered by downlink slots in other cells. However, that kind of interference scenarios
are visible in unsynchronized TDD deployments, for example when dynamic TDD with different UL:DL ratios
have been applied in neighbouring cells. In all, four different inter-operator interference scenarios can be
listed, as shown in Fig. 1:
Interference from base station to mobile terminal (downlink-to-downlink interference), valid for both synchronized and unsynchronized TDD deployments as well as for frequency division duplex (FDD) deployments. This will impact mostly the users located closest to the interfering base stations, for example the users located next to the outer walls of the buildings. Since the transmission power of a
base station is typically considerably higher than the average transmission power of a power-controlled mobile terminal, this interference scenario is estimated to be highly critical for the assumed multi-operator scenario.
Interference from mobile terminal to base station (uplink-to-uplink interference), valid for both synchronized and unsynchronized TDD deployments as well as for FDD deployments. Due to the fact that the typical transmission power of a mobile terminal is quite low, in particular if the mobile is served by a small cell, and that the interfering mobile terminal is located quite far from the victim base station, this kind of interference scenario is estimated to be less critical.
Interference from base station to base station (downlink-to-uplink interference), valid for unsynchronized TDD deployments. Due to the higher transmission power, and higher antenna gain, this kind of interference scenario is more critical than the previous one, and should be evaluated more carefully. However, the impact of the interference on the uplink performance can potentially be reduced by adjusting the uplink power control settings within the victim cell as discussed in [16].
Interference from mobile terminal to mobile terminal (uplink-to-downlink interference), valid for unsynchronized TDD deployments. This kind of interference scenario can be seen as less critical, due to the lower transmission power levels and the fact that the interferer and the victim are typically not located that close to each other.
The motivations listed above may differ for different types of multi-operator deployment scenarios, for
example if the micro operators are located within the same building. Therefore, for a complete interference
analysis, the impact of all four inter-operator interference scenarios should be considered. However, in this
paper the focus is only on the first one, i.e., on the scenario where the transmissions from base stations
belonging to one micro operator (uO2) are interfering the downlink reception of a mobile terminal belonging
to another micro operator (uO1). This interference scenario is selected as it is estimated to represent the
most critical type of inter-operator interference for the assumed deployment scenario. This is motivated both
by the higher transmission power of a base station compared to a mobile terminal, and by the higher receiver
sensitivity of a mobile terminal compared to a pico base station.
Fig. 1 Different inter-operator interference scenarios between neighboring micro operators
DLDLUL
uO1 uO2
Desired signal
Interfering signal
UL
ULDLDL UL
uO1:
uO2:
Synchronized slotUnsynchronized slot
3 System model This section provides a brief description of the assumed network layout for a scenario with two neighbouring
micro operators. Furthermore, the applied propagation models, both for the indoor propagation within the
micro operator building and for the building-to-building propagation between the micro operator buildings
are presented. Finally, a model for the user performance, i.e., the average user throughput, is discussed.
3.1 Network layout The evaluated micro operator network deployment model consists of two equally-sized buildings (50 x 120
m, based on the indoor deployment scenario defined in [17]), located at a distance 𝐷 from each other, see
Fig. 2. In this initial study, the buildings are assumed to be in line-of-sight (LOS) with each other and only one
floor per building is modelled. Micro operator 1 (uO1) is assumed to be serving users within the first building,
while micro operator 2 (uO2) is serving users within the second building. Finally, it is assumed that uO2 has
deployed 12 pico base stations per floor, while the density of pico base stations belonging to uO1 is varied
between 1 and 12 pico base stations per floor. In case of 12 base stations per floor, the base station locations
are the same as defined in [17], and for the lower base station densities the base stations have been deployed
so that a roughly uniform coverage can be obtained throughout the floor area. In this initial study focusing
on the downlink performance, mobile terminals are modelled only within the uO1 building, assuming a
uniform user density over the floor area.
Fig. 2 Description of the assumed network layout with two micro operators, uO1 and uO2
3.2 Propagation models When evaluating the impact of interference between neighbouring micro operators, path losses between
the different nodes (base stations and mobile terminals) have to be estimated first. Hence, appropriate wave
T
R
D
T1
T2
T3
T4
R1
R2
R3
R4
LOS path
NLOS paths
NLOS paths
uO1 uO2
propagation models for both the desired links within the building (indoor wave propagation) and the
interfering links between the buildings (building-to-building wave propagation) are needed. In the
propagation model proposed in this paper, the coupling loss 𝐶𝑚,𝑏 between a mobile terminal 𝑚 and a base
station 𝑏 is calculated as
𝐶𝑚,𝑏,𝑑𝐵 = 𝐿𝑚,𝑏,𝑑𝐵 − 𝐺𝐵𝑆,𝑚,𝑏,𝑑𝐵 − 𝐺𝑀𝑇,𝑚,𝑏,𝑑𝐵 + 𝑋𝑚,𝑏,𝑑𝐵 (1)
where 𝐺𝐵𝑆,𝑚,𝑏 and 𝐺𝑀𝑇,𝑚,𝑏 are the base station and mobile terminal antenna gains, respectively, and 𝑋𝑚,𝑏 is
a log-normally distributed random value modelling the impact of shadow fading. Next, the assumed
modelling of the path loss 𝐿𝑚,𝑏 is discussed in more detail.
The indoor propagation is modelled using the 3GPP Indoor - Mixed Office propagation model as defined in
[17]. The model includes both a LOS (𝐿𝐿𝑂𝑆) and a non-line-of-sight (NLOS) component (𝐿𝑁𝐿𝑂𝑆), and the LOS
probability (𝑃𝐿𝑂𝑆) is defined to decrease as a function of the distance between the base station and the
mobile terminal. The standard deviation of the shadow fading is assumed to be equal to 3 dB (LOS) or 8 dB
(NLOS). Furthermore, both the shadow fading and the LOS probability are spatially correlated, assuming
correlation distances equal to 10 m or 6 m (shadow fading in LOS or NLOS), and 10 m (LOS probability) [17].
While there are several different types of indoor propagation and building penetration models described in
the literature, building-to-building propagation models have received much less attention. One such model
is the 3GPP dual-strip model described in [14]. Unfortunately, the 3GPP dual-strip model has its weaknesses,
as demonstrated also by the measurements presented in [18]: it is tailored for 2 GHz or 3.5 GHz and is not
valid for higher frequencies, the building penetration model considers only one of the building walls, the
impact of both the building wall material and the angle of incidence on the building penetration loss is not
taken into account, and the outdoor path loss between the buildings is to a large extent based on a macro
cellular NLOS model even at relatively short distances.
In order to consider the weaknesses of using the 3GPP dual-strip model in the micro operator deployment
scenario, a new type of building-to-building propagation model has been defined for this study. The overall
model is composed of a number of different components taking into account free space propagation in LOS,
diffraction modelling in NLOS, building penetration loss and indoor loss. The basis for each of these
components has been taken by selecting appropriate models described in the literature. These include the
recursive microcell model for outdoor propagation [19-20], the general building penetration model [20], the
model for LOS building penetration [20-22] including the impact of the angle of incidence, the model for
NLOS building penetration [17] and the linear attenuation model for indoor propagation [17]. Furthermore,
the building wall loss is assumed to depend both on the frequency (𝑓𝑐) and on the wall material as defined in
[17].
The main principle of determining the path loss from an indoor node to another indoor node located in the
neighbouring building is described in Fig. 2. For both buildings, four different sub-paths, one through each
outer wall, are evaluated [20]. Hence, for each link between a transmitter and a receiver, the total received
power is calculated as a linear sum of the received powers from all the 16 different sub-paths. Each sub-path
takes into account both the outdoor loss between the outer wall reference points (T1-T4 and R1-R4 in Fig.
2), and the building penetration and indoor losses for both buildings. In all, the path loss per sub-path is
calculated as
𝐿𝑑𝐵 = 𝐿𝑖𝑛,1 + 𝐿𝑜𝑤,1(𝑓𝑐) + 𝐿𝑜𝑢𝑡(𝑓𝑐) + 𝐿𝑜𝑤,2(𝑓𝑐) + 𝐿𝑖𝑛,2 (2)
In (2), 𝐿𝑖𝑛 is the indoor loss, modelled as 𝐿𝑖𝑛 = 0.5𝑑2𝐷−𝑖𝑛, where 𝑑2𝐷−𝑖𝑛 is the two-dimensional distance
between the indoor node and the outer wall reference point [17]. Parameter 𝐿𝑜𝑤 models the building wall
loss and it consists of two components: one that depends on the angle of incidence 𝜃 and the other that
depends on the wall material and center frequency 𝑓𝑐. Depending on the desired building penetration model
(LOS or NLOS), 𝐿𝑜𝑤 is calculated either as described in (3) [22] or in (4) [17].
𝐿𝑜𝑤,𝐿𝑂𝑆 = 20(1 − cos 𝜃)2 + 𝐿𝑚𝑎𝑡𝑒𝑟𝑖𝑎𝑙(𝑓𝑐) (3)
𝐿𝑜𝑤,𝑁𝐿𝑂𝑆 = 5 + 𝐿𝑚𝑎𝑡𝑒𝑟𝑖𝑎𝑙(𝑓𝑐) (4)
Since a LOS propagation condition is assumed between the buildings, 𝐿𝑜𝑤 is based on (3) for wall reference
points T3 and R1, while it is based on (4) for all the other wall reference points. In higher frequency bands,
the material characteristics of the building have a major impact on defining the penetration loss pattern,
which means that the value of 𝐿𝑚𝑎𝑡𝑒𝑟𝑖𝑎𝑙(𝑓𝑐) can be significantly different for different buildings, see for
example the discussion and measurement results in [17, 22-25]. In this paper, 𝐿𝑚𝑎𝑡𝑒𝑟𝑖𝑎𝑙(𝑓𝑐) follows the
model in [17], i.e., two different frequency-dependent wall loss values have been assumed: low-loss (old
buildings with 30% standard multi-pane windows, and 70% concrete) and high-loss (modern buildings with
70% infrared reflective (IRR) glass windows, and 30% concrete).
Parameter 𝐿𝑜𝑢𝑡 in (2) is the outdoor path loss between the outer wall reference points. In case of a LOS path
(sub-path between T3 and R1 in Fig. 2), 𝐿𝑜𝑢𝑡 is based on a free space propagation model, and the applied
distance is the sum of the outdoor and the indoor distances [20, 21]. In case of NLOS paths, 𝐿𝑜𝑢𝑡 considers
only the path loss between the outer wall reference points, and the path loss is based on the recursive
microcell model [19], assuming a breakpoint for the path loss exponent at 300 m. Finally, when it comes to
the shadow fading model for the building-to-building penetration, standard deviation equal to 6 dB and
correlation distance equal to 10 m are assumed.
While the main assumption in this paper is that the buildings are in LOS with each other, it is also possible to
model a NLOS propagation condition for the outdoor environment surrounding the buildings. For that case,
the proposed building-to-building model can be modified by replacing the recursive microcell model for
example with the 3GPP Urban Micro – Street Canyon model defined in [17], and applying the NLOS building
penetration model for all wall reference points. Furthermore, standard deviation of the shadow fading is
increased to 8 dB and the correlation distance is increased to 13 m [17].
An example of the resulting received signal strength (RSS) heat maps is visualized in Fig. 3 for a scenario with
low-loss building walls, center frequency equal to 3.5 GHz, distance 𝐷 equal to 50 m, isotropic base station
and mobile terminal antennas, and base station transmission power equal to 24 dBm. The impact of shadow
fading has not been taken into account, which means that the different colours in Fig. 3 illustrate the median
RSS values for each location. As can be seen by looking at the building on the left (uO1), one pico base station
deployed in the middle of the floor is sufficient to provide RSS higher than -70 dBm throughout the
investigated area. A denser deployment improves the coverage, and therefore, the median RSS within the
building on the right (uO2) is better than -50 dBm for all locations. When it comes to the total downlink
interference from uO2 to uO1, the highest interference levels (approximately equal to -70 dBm) can be found
next to the wall facing the neighbouring building. Furthermore, it is clearly visible that the inter-operator
interference gets weaker when moving farther away from the illuminated wall. One can also notice that at
the upper and lower right corner of the uO1 building the level of interference is quite close to the RSS from
the serving base station, which suggests that for those locations the impact of inter-operator interference on
the user performance will be quite dramatic. At the same time the downlink interference from uO1 to uO2
is much weaker, which is due to fact that the interference is caused by just one pico base station located
much farther away from the building wall compared to uO2.
Fig. 3 Received signal strength for the desired signal (figure on the left) and for the total inter-operator interference (figure on the right)
3.3 User performance The main output from the evaluations is the impact of inter-operator interference on the user performance,
or more specifically, on the average user throughput. Generally speaking, average user throughput takes into
account both the impact of scheduling and link adaptation. Furthermore, one key input for the link
adaptation is the received signal-to-interference-and-noise-ratio (SINR).
The downlink SINR for mobile terminal 𝑚, being served by base station 𝑏, can be calculated as
𝛾𝑚,𝑏 =𝑃𝑚,𝑏𝐶𝑚,𝑏
−1
𝐼𝑜𝑤𝑛,𝑚,𝑏 + 𝐼𝑜𝑡ℎ𝑒𝑟,𝑚 +𝑁𝑚=
𝑃𝑚,𝑏𝐶𝑚,𝑏−1
∑ 𝛼𝑗𝑃𝑗𝐶𝑚,𝑗−1
𝐵
𝑗=1𝑗≠𝑏
+∑ 𝛼𝑘𝑃𝑘𝜌𝑘−1𝐶𝑚,𝑘
−1𝐾
𝑘=1+𝑁𝑚
(5)
In (5), 𝐼𝑜𝑤𝑛 is the received power from all the other base stations belonging to the serving micro operator
(having a total of 𝐵 base stations), while 𝐼𝑜𝑡ℎ𝑒𝑟 is the received power from all the base stations belonging to
the other micro operators (a total of 𝐾 base stations). Furthermore, 𝑃𝑚,𝑏 is the transmission power from
base station 𝑏 towards mobile terminal 𝑚, 𝐶𝑚,𝑏 is the coupling loss between base station 𝑏 and mobile
terminal 𝑚, 𝑁𝑚 is the thermal noise power of mobile terminal 𝑚, and 𝜌𝑘 is the adjacent channel interference
ratio (ACIR) taking into account the appropriate adjacent channel attenuation between the micro operators
when applicable [26]. Finally, the activity factor 𝛼𝑗 is equal to 1 when base station 𝑗 is transmitting at the
same time and on the same frequency resources as base station 𝑏, otherwise 𝛼𝑗 is equal to 0.
For simplicity, only time-domain scheduling of users is assumed, which means that when mobile terminal 𝑚
is scheduled, 𝑃𝑚,𝑏 becomes equal to the total base station output power (𝑃𝐵𝑆). Furthermore, when defining
the coupling loss towards the serving base station (𝐶𝑚,𝑏), the user is always assumed to be served by the
base station, which has the smallest coupling loss towards the mobile terminal.
The SINR is mapped to a bit rate using the Shannon-Hartley theorem:
𝜏𝑚,𝑏 = 𝛽 log2(1 + 𝛾𝑚,𝑏) (6)
where 𝛽 is the bandwidth of the channel. Finally, assuming a Round Robin scheduling algorithm, and that
base station 𝑏 is simultaneously serving a total of 𝑢𝑏 users, the average user throughput becomes equal to
𝑅𝑚,𝑏 =𝜏𝑚,𝑏𝑢𝑏
=𝛽 log2(1 + 𝛾𝑚,𝑏)
𝑢𝑏 (7)
This initial study investigates the impact of inter-operator interference on the user performance within a
scenario, where the victim micro operator (uO1) is serving only one active user, i.e., the activity factor 𝛼𝑗 in
(5) is equal to zero for all the other base stations (𝐼𝑜𝑤𝑛 = 0), and parameter 𝑢𝑏 in (7) is equal to one. At the
same time, the interfering micro operator network (uO2) is assumed to be fully loaded, i.e., 𝛼𝑘 = 1 ∀𝑘.
Taking all this into account, the average user throughput can now be written as
𝑅𝑚,𝑏 = 𝛽 log2
(
1+𝑃𝐵𝑆𝐶𝑚,𝑏
−1
𝑁𝑚 +𝑃𝐵𝑆𝜌∑ 𝐶𝑚,𝑘
−1𝐾
𝑘=1 )
(8)
4 Evaluation results In order to evaluate the impact of the inter-operator interference from a micro operator network inside a
building to another micro operator network in a neighbouring building, system simulations have been carried
out using the system and performance models from Section 3. The main simulation parameters are listed in
Table I. It should be noted that the ACIR parameter takes into account both the adjacent channel leakage
power ratio (ACLR) of the transmitter, assumed to be equal to 45 dB [27], and the adjacent channel selectivity
(ACS) of the receiver, equal to 27 dB [28].
During the Monte-Carlo simulations, the evaluated victim user is dropped in random positions inside the uO1
building and coupling loss values are calculated towards all the base stations within both buildings. Based on
the obtained coupling losses, the serving base station is defined, and the level of the total inter-operator
interference (𝐼𝑜𝑡ℎ𝑒𝑟) is calculated. Finally, the SINR and the corresponding average user throughput are
obtained.
Table I List of the main simulation parameters
Parameter Value
Center frequency (𝑓𝑐) 3.5 GHz
Channel bandwidth (𝛽) 20 MHz
Base station transmission power (𝑃𝐵𝑆) 24 dBm
Mobile terminal noise power (𝑁) -92.4 dBm
Adjacent channel interference ratio (𝜌) 0 dB (co-channel) 26.9 dB (adjacent channel)
Base station antenna gain (𝐺𝐵𝑆) 5 dBi (omnidirectional)
Mobile terminal antenna gain (𝐺𝑀𝑇) 0 dBi (omnidirectional)
Building wall loss (𝐿𝑚𝑎𝑡𝑒𝑟𝑖𝑎𝑙) 7.7 dB (low-loss wall) 21.8 dB (high-loss wall)
A summary of the simulation results with different deployment options is shown in Fig. 4, Fig. 5 and Fig. 6,
where the received total inter-operator interference (𝐼𝑜𝑡ℎ𝑒𝑟), the average user performance (calculated as
the average over the obtained 𝑅𝑚,𝑏 values) and the performance of the worst 5th percentile are studied,
while increasing the distance 𝐷 between the buildings from 10 m to 900 m. The inter-operator interference
is evaluated by calculating the probability that the level of 𝐼𝑜𝑡ℎ𝑒𝑟 is higher than 𝑁 − 6 dBm, which has quite
often been applied as the threshold for harmful interference, see for example [29]. Furthermore, the
downlink throughput loss values in Fig. 5 and Fig. 6 have been obtained by comparing the user throughput
values of the multi-operator scenario to the corresponding values within the single operator scenario (when
𝐼𝑜𝑡ℎ𝑒𝑟 = 0).
Fig. 4 Probability that the received total inter-operator interference is higher than the threshold for harmful interference
Fig. 5 Average downlink throughput loss within the uO1 building as a function of the distance between the buildings
Fig. 6 Downlink throughput loss for the worst 5th percentile of users within the uO1 building as a function of the distance between the buildings
In order to set the limit for the maximum allowed performance degradation, the approach taken in [26] can
be adopted. Thus, in this paper the minimum separation distance 𝐷𝑚𝑖𝑛 between the two buildings is defined
so that the average user performance is not degraded more than 1%, and at the same time the performance
of the worst 5th percentile is not degraded more than 5%.
Looking at the curves in Fig. 4, 𝐼𝑜𝑡ℎ𝑒𝑟 seems to exceed the threshold for harmful interference with a very
large probability when a co-channel deployment with low-loss building walls is assumed. This corresponds
also nicely to the findings in Fig. 5 and Fig. 6, which show that the required 𝐷𝑚𝑖𝑛 becomes as large as 830 m.
Furthermore, if the required 𝐷𝑚𝑖𝑛 is compared to the results in Fig. 4, one can notice that it matches roughly
to the situation when 𝐼𝑜𝑡ℎ𝑒𝑟 exceeds the threshold for harmful interference with 10% probability.
In case of the other deployment scenarios, the level of 𝐼𝑜𝑡ℎ𝑒𝑟 is much lower, exceeding the interference
threshold in less than 20% of the time even when the buildings are located close to each other. In all, the
required 𝐷𝑚𝑖𝑛 values become equal to: 80 m (high-loss, co-channel), 120 m (low-loss, adjacent channel) and
less than 10 m (high-loss, adjacent channel). Again, if these values are compared to the curves in Fig. 4, they
seem to match quite well the results when 𝐼𝑜𝑡ℎ𝑒𝑟 exceeds the interference threshold with 10% probability.
When the buildings are in NLOS with each other, the level of 𝐼𝑜𝑡ℎ𝑒𝑟 is attenuated much faster, and the
required 𝐷𝑚𝑖𝑛 becomes considerably smaller as can be seen in the results shown in Fig. 7. The required 𝐷𝑚𝑖𝑛
is now equal to 180 m for the co-channel deployment with low-loss walls, and less than 50 m for all the other
deployment scenarios.
Fig. 7 Downlink throughput loss within the uO1 building as a function of the distance between the buildings. The buildings are assumed to be in NLOS with each other
The final thing to study is the impact of the base station density within the victim building on the required
𝐷𝑚𝑖𝑛. With the traditional approach based on 𝐼𝑜𝑡ℎ𝑒𝑟, there would not be any difference, but with the
proposed approach based on the user performance, some kind of impact could be expected. Generally
speaking, when the base station density is increased, coupling loss from the mobile terminal towards the
serving base station is reduced, improving the signal-to-noise-ratio (SNR). Since the users, in particular the
worst ones, experience now much higher bit rates (users are higher up on the Shannon-Hartley curve), a
certain 𝐼𝑜𝑡ℎ𝑒𝑟 will then have a smaller relative impact on the user performance compared to the users
experiencing lower SNR values. Alternatively, the victim users can tolerate a higher 𝐼𝑜𝑡ℎ𝑒𝑟 for a certain level
of performance degradation. However, the improvement may not be as large from the average system
performance point of view.
Looking at the curves in Fig. 8 for the adjacent channel deployment with low-loss building walls this indeed
seems to be the case. When the base station density is increased from 1 to 12 pico base stations per floor,
the required 𝐷𝑚𝑖𝑛 is reduced from 120 m to 60 m. Furthermore, it is very clear that the 𝐷𝑚𝑖𝑛 becomes more
and more limited by the average performance and not by the performance of the worst users. A very similar
conclusion can be drawn also for the co-channel deployment with low-loss building walls, where the required
𝐷𝑚𝑖𝑛 is reduced from 830 m to 710 m. Finally, from the 𝐼𝑜𝑡ℎ𝑒𝑟 statistics point of view, the maximum tolerable
probability of harmful interference is at the same time increased from 10% to 18%, as indicated by the results
in Fig. 4.
Fig. 8 Downlink throughput loss within the uO1 building as a function of the distance between the buildings, assuming different base station densities
5 Conclusions and further work Local small cell deployments will become an important part of the future 5G networks, in particular for the
deployments within higher frequency bands. In order to make such ultra-dense network deployments more
cost-efficient, new business and spectrum authorization models are needed in addition to the traditional
models based on networks deployed and operated by the MNOs. One such model is the recently introduced
concept for 5G micro operators for allowing different stakeholders to deploy and operate their own networks
in specific buildings using locally issued micro licensed spectrum with pre-defined level of quality guarantees.
One important aspect of the proposed concept is the required minimum separation distance between the
micro operators in co-channel and adjacent channel deployments so that the level of inter-operator
interference is still sufficiently low. To characterize such interference, and in particular the impact of the
interference on the performance of the victim network, models for appropriate micro operator network
deployments, radio wave propagation and user performance are needed. To support such interference
evaluations, this paper has proposed a deployment scenario including two neighbouring buildings,
propagation models for connections both within a building and between the buildings, and a criteria for the
required minimum separation distance based on the observed throughput loss. Finally, system simulations
have been performed to evaluate the impact of the key deployment aspects on the required minimum
separation distance between the micro operators including building type, spectrum allocation, surrounding
outdoor environment and base station density.
The obtained results in the case of 3.5 GHz band indicate that if the building wall loss is low (old buildings
with traditional windows) and the buildings are in line-of-sight with each other, the level of inter-operator
interference becomes high, and a large minimum separation distance (up to 830 m) is required if the micro
operators are operating on the same frequency channel. In case of modern buildings with modern infrared
reflective windows the building wall loss becomes much higher, and hence, the required minimum separation
distance is reduced to 80 m. Furthermore, if the operators are assigned adjacent channels, the minimum
separation distance becomes equal to 120 m (low-loss building walls) or less than 10 m (high-loss building
walls). If the buildings are not in line-of-sight with each other, the outdoor environment surrounding the
buildings will attenuate the interference much faster, and thus, the required minimum distance is reduced
to 180 m even for a scenario with a co-channel deployment and low-loss building walls. Finally, the impact
of the inter-operator interference is shown to depend on the base station density within the victim building:
the required minimum separation distance is reduced from 830 m to 710 m (co-channel, low-loss building
walls) and from 120 m to 60 m (adjacent channel, low-loss building walls) when the base station density
within the victim building is increased from 1 to 12 base stations per floor.
As a summary, the obtained results suggest that both the maximum tolerable level of inter-operator
interference (𝐼𝑜𝑡ℎ𝑒𝑟) and the required minimum separation distance between two micro operators will be
highly scenario-specific. Therefore, the traditional approach of defining a certain fixed interference threshold
dimensioned for the worst case scenario, is no longer feasible for shared spectrum access as it results in less
efficient reuse of the spectrum. This will need to be taken into account in the development of future 5G
spectrum authorization models, such as the new micro licensing model, and the setup of appropriate rules
to define and coordinate the resulting interferences. Future research is therefore needed to characterize the
inter-operator interference even further. For example, the impact of center frequency, base station density
and the base station transmission power of the interfering network, traffic load within the victim network
(i.e., the impact of 𝐼𝑜𝑤𝑛), more advanced interference coordination mechanisms, beamforming, and
signalling capabilities between the micro operators should be investigated. Furthermore, it would be valuable
to include also the other interference scenarios, in addition to the downlink-to-downlink interference
discussed in this paper, into the overall interference analysis. Finally, the analysis could be extended also to
a deployment scenario, where the micro operators are located within the same building.
References 1. EC. (2016). Communication from the Commission to the Parliament, the Council, the European
Economic and Social Committee and the Committee of the Regions. Connectivity for a competitive digital single market – Towards a European gigabit society. COM (2016) 587 Final. European Commission.
2. 5GPPP. (2017). 5G innovations for new business opportunities. The 5G Infrastructure Public Private Partnership.
3. Matinmikko, M., Latva-aho, M., Ahokangas, P., Yrjölä, S. & Koivumäki, T. (2017). Micro operators to boost local service delivery in 5G. Wireless Personal Communications, 95(1), 69-82.
4. Zander, J. (2017). Beyond the ultra-dense barrier: Paradigm shifts on the road beyond 1000x wireless capacity. IEEE Wireless Communications, 24(3), 96-102.
5. Saldana, J., et al. (2017). Alternative networks: Toward a global access to the Internet for all. IEEE Communications Magazine, 55(9), 187-193.
6. EC. (2016). 5G for Europe: An action plan. Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions. COM (2016) 588 Final. European Commission.
7. CEPT ECC. (2014). Licensed shared access. ECC Report 205. European Conference of Postal and Telecommunications, Electronic Communications Committee.
8. FCC. (2016). Amendment of the Commission’s Rules with Regard to Commercial Operations in the 3550-3650 MHz Band, Order of Reconsiderations and Second Report and Order. FCC-16-55. Federal Communications Commission.
9. Matinmikko, M., Roivainen, A., Latva-aho, M. & Hiltunen, K. (2017). Interference study of micro licensing for 5G micro operator small cell deployments. Conference on Cognitive Radio Oriented Wireless Networks and Communications (CrownCom), Lisboa, Portugal, 20-21 September 2017.
10. Matinmikko, M., Latva-aho, M., Ahokangas, P. & Seppänen, V. (2018). On regulations for 5G: Micro licensing for locally operated networks. Telecommunications Policy (to appear).
11. Ahokangas, P., Moqaddamerad, S., Matinmikko, M., Abouzeid, A., Atkova, I. Francis Gomes, J., & Iivari, M. (2016). Future micro operators business models in 5G. In International conference on Restructuring of the Global Economy (ROGE).
12. Gupta, A. K., Andrews, J. G. & Heath, R. W. (2016). On the feasibility of sharing spectrum licenses in mmWave cellular systems. IEEE Transactions on Communications, 64(9), 3981-3995.
13. Gupta, A. K., Alkhateeb, A., Andrews, J. G. & Heath, R. W. (2016). Gains of restricted secondary licensing in millimeter wave cellular systems. IEEE Journal on Selected Areas in Communications, 34(11), 2935-2950.
14. 3GPP. (2013). Small cell enhancements for E-UTRA and E-UTRAN – Physical layer aspects. 3GPP TR 36.872, V12.1.0. 3rd Generation Partnership Project.
15. Parkvall, S., Dahlman, E., Furuskär, A. & Frenne, M. (2017). NR: The new 5G radio access technology. IEEE Communications Standards Magazine. 1(4), 24-30.
16. Hiltunen, K. (2014). The performance of dense and heterogeneous LTE network deployments within an urban environment. Doctoral dissertation. Aalto University.
17. 3GPP. (2017). Study on channel model for frequencies from 0.5 to 100 GHz. 3GPP TR 38.901, V14.1.1. 3rd Generation Partnership Project.
18. Lehtomäki, J., Vuohtoniemi, R., Matinmikko-Blue, M. & Hiltunen, K. (2018). Building-to-building propagation loss measurements at 3.5 GHz with application to micro operators. In IEEE Wireless Communications and Networking Conference (to appear).
19. Berg, J.-E. (1995). A recursive method for street microcell path loss calculations. In IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC).
20. Damosso, E. & Correia, L. M., Eds. (1999). Digital mobile radio towards future generation systems. Final report. European Commission. http://www.lx.it.pt/cost231/final_report.htm. Accessed 15 May 2017.
21. Berg, J.-E. (1996). Building penetration loss along street microcells. In IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC).
22. Semaan, E., Harrysson, F., Furuskär, A. & Asplund, H. (2014). Outdoor-to-indoor coverage in high frequency bands. In IEEE Globecom 2014 Workshop – Mobile Communications in Higher Frequency Bands.
23. Ökvist, P., Asplund, H., Simonsson, A., Halvarsson, B., Medbo, J. & Seifi, N. (2015). 15 GHz propagation properties assessed with 5G radio access prototype. In IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC) – Workshop on 5G Channel Measurements and Modeling.
24. ITU-R. (2016). Compilation of measurement data relating to building entry loss. Report ITU-R P.2346-1. International Telecommunication Union Radiocommunication sector.
25. Aalto University, AT&T, BUBT, CMCC, Ericsson, Huawei, Intel, KT Corporation, Nokia, NTT DOCOMO, New York University, Qualcomm, Samsung, University of Bristol, University of Southern California. (2015). 5G channel model for bands up to 100 GHz. White paper. http://www.5gworkshops.com/5GCM.html. Accessed 15 May 2017.
26. 3GPP. (2017). Radio Frequency (RF) system scenarios. 3GPP TR 36.942, V14.0.0. 3rd Generation Partnership Project.
27. 3GPP. (2017). Base station (BS) radio transmission and reception. 3GPP TS 36.104, V14.4.0. 3rd Generation Partnership Project.
28. 3GPP. (2017). User equipment (UE) radio transmission and reception. 3GPP TS 36.101, V14.4.0. 3rd Generation Partnership Project.
29. CEPT ECC. (2012). Broadband wireless systems usage in 2300-2400 MHz. ECC Report 172. European Conference of Postal and Telecommunications, Electronic Communications Committee